Tumor invasion and associated angiogenesis evoke a remodeling of extracellular matrix components. Retroviral vectors bearing auxiliary matrix-targeting motifs (i.e., collagen-binding polypeptides) accumulate at sites of newly exposed collagen, thus promoting tumor site-specific gene delivery. In this study, we assessed the antitumor effects of serial portal vein infusions of matrix-targeted vectors bearing a mutant cyclin G1 (dnG1) construct in a nude mouse model of liver metastasis. The size of tumor foci was dramatically reduced in dnG1 vector-treated mice compared with that in control vector- or PBS-treated animals (P = 0.0002). These findings represent a definitive advance in the development of targeted injectable vectors for metastatic cancer.

Retroviral vectors are currently used in over 70% of human cancer gene therapy trials: more than 3000 patients have received a gene therapy vector in over 300 approved Phase I/II gene therapy protocols(1). Thus far, it appears that retroviral vectors are relatively safe and have minimal risk of insertional mutagenesis or oncogenesis (2). For cancer gene therapy, a built-in safety feature provided by retroviral vectors is the selective transduction of actively dividing cells (3), thus sparing normal nondividing cells. Retroviral transduction of target cells is initiated by viral binding to cell surface receptors (4, 5), followed by fusion of viral and cellular membranes(6) and core entry. Important advances in the production(7, 8) and physiological stability (9) of retroviral vectors have been made in recent years. However, the potential use of these vectors for in vivo gene delivery is daunted by logistical problems, i.e., limited retroviral bioavailability due to lack of tissue specificity (1, 10). Thus, the development of targeted injectable retroviral vectors for human gene therapy has been hitherto elusive (1, 10). Current research on retroviral targeting has focused on molecular engineering of the retroviral envelope (env) proteins(11, 12, 13). In contrast to various strategies involving the display of polypeptide ligands (11, 14) or single-chain antibodies (15, 16, 17) on viral envelopes, we undertook the seemingly counterintuitive strategy of targeting the ECM3itself, i.e., targeting ECM proteins exposed in pathophysiological lesions to concentrate vectors in the ECM in the vicinity of target cells (18, 19). For transduction of human cells, an optimized targeting construct was engineered to display a matrix-targeting motif on the NH2-terminal region of the amphotropic 4070A envelope (20). Resultant single enveloped vectors (which would be readily amenable to large-scale clinical vector production) exhibited both ECM targeting and WT infectivity. We subsequently demonstrated successful transduction of human cancer cells in vivo by portal vein infusions of the matrix-targeted vector but not the nontargeted vector in a nude mouse model of liver metastasis (20). In the present study, we report dramatic reductions in the sizes of metastatic tumor foci in nude mice by portal vein infusions (via an indwelling catheter) of matrix-targeted retroviral vectors bearing a cytocidal mutant cell cycle control gene.

Cells, Cell Culture Conditions, Plasmids and Vectors Bearing Marker, and Cell Cycle Control Genes.

NIH3T3, 293T, and human pancreatic cancer MiaPaca2 cells were supplied by American Type Culture Collection. NIH3T3 and 293T cells were maintained in DMEM supplemented with 10% fetal bovine serum (D10;Biowhittaker). The pcgp plasmids containing the viral gag pol genes and a retroviral vector, pcnBg, expressing a nucleus-targeted β-galactosidase construct were kindly provided by Drs. Paula Cannon and Ling Li, respectively (University of Southern California Gene Therapy Laboratories, Los Angeles, CA). The plasmid containing VSVG env protein was kindly provided by Dr. Theodore Friedmann (University of California, San Diego, CA). A truncated (amino acids 41–249) cyclin G1 construct (dnG1) was cloned into the retroviral expression vector pREX.

Production of Matrix-targeted Retroviral Vectors Bearing Mutant Cyclin G1 Constructs.

High-titer vectors were generated using a transient three- or four-plasmid cotransfection system (7) in which the packaging components gag-pol and a chimeric murine leukemia virus-based env bearing a vWF-derived collagen-binding (matrix-targeting) motif expressed from the cytomegalovirus promoter were placed on separate plasmids, each of which contained the SV40 origin of replication. The vectors expressed without WT env were named Bv1 or Hs2 (Bv, bovine vWF-derived; Hs, human vWF-derived; LF or 1, linkers derived from natural vWF sequences; LS or 2, standard linkers). To further increase viral titer, a fusogenic VSVG env protein(21) was coexpressed with Bv1 or Hs2 envproteins in a four-plasmid cotransfection protocol.

Viral titers in murine NIH3T3 cells were determined as described previously, based on expression of the β-galactosidase or neomycin phosphotransferase resistance (neor) gene(22). Viral titer was expressed as the number of G418-resistant cfu/ml and ranged from 106−108 cfu/ml, depending on the nature and amount of plasmid DNA used in the transfection protocol.

In Vitro Efficacy Studies.

To assess the cytocidal/cytostatic effects of the dnG1 vector, the transduced cells were evaluated for their proliferative potential by counting the number of viable cells in each culture at serial intervals(up to 4 days) after transduction without G418 selection. Western analysis of endogenous and mutant cyclin G1 protein expression was performed as described previously (22).

In vivo efficacy studies were conducted in compliance with a protocol approved by the University of Southern California Institution Animal Care and Use Committee. To evaluate the efficiency of targeted gene delivery based on the antitumor effects of dnG1 vector treatment in vivo, a model of liver metastasis simulating the route of dissemination of human colon cancer was established in nude mice. Briefly, 7 × 105 tumor cells were infused slowly into the portal vein via an indwelling catheter that was kept in place for 14 days. Intracatheter infusions of either a low- or high-dose dnG1 vector (titers, 3 × 106 or 9 × 108 cfu/ml at 200 μl/day), a high-dose control vector bearing a β-galactosidase gene (titer, 8.5 × 108 cfu/ml), or PBS (pH 7.4) were begun 3 days later and continued daily for a total of 9 days. The mice were sacrificed by cervical dislocation 1 day after completion of treatment.

Histological and Morphometric Analysis.

The liver lobes were excised, fixed in 10% formalin, labeled (A,right and caudate lobes; B, left lobe; C, median lobe), processed separately, and embedded in paraffin blocks. The antitumor efficacy of dnG1 vector treatment was assessed as follows: H&E-stained tissue sections were examined by light microscopy; and the SAs of representative liver sections and tumor foci from lobes A, B, and C were measured by morphometric analysis using an Optimas image analysis system. Evaluation of retroviral safety included assessment of the integrity of the liver architecture and examination for the presence of hepatocellular swelling or necrosis, inflammatory infiltrates,cholestasis, and/or thrombosis. Tissue sections were also immunostained to assess the presence of β-galactosidase transgene(20), cytokeratin, apoptosis, PAS, vimentin, and CD68.

Statistical Analysis.

For the in vivo efficacy study, four treatment groups were compared: (a) low-dose Bv1/dnG1 (titer, 3 × 106 cfu/ml); (b) high-dose Hs2/VSVG/dnG1 (titer, 9.5 × 108cfu/ml); (c) high-dose control vector bearing aβ-galactosidase (nBg) gene (titer, 8.5 × 108 cfu/ml); and (d) PBS control. Initially, 12 mice were studied; 4 were treated with a high-dose dnG1 vector, 4 were treated with a high-dose control vector, and 4 were treated with PBS. Subsequently, four additional mice were treated with a low-dose dnG1 vector. The response variables (total SA of the liver,total SA of the tumor, tumor SA:liver SA ratio, and mean SA of tumor foci) were log-transformed before formal analysis. A repeated measures analysis with lobe as the repeated measures factor was used to determine whether or not the treatment had an effect on each of the response variables. Pairwise comparisons were also performed for the outcome variables with overall F test Ps of <0.05 between groups.

In 1994, our laboratory cloned a human G-type cyclin(CYCG1), which was markedly overexpressed in a subset of osteosarcoma cells (23). Since then, the retroviral transfer of an antisense cyclin G1 construct has been repeatedly shown to induce down-regulation of cyclin G1 expression and apoptosis of target cells, inhibit proliferation of numerous cancer cells in vitro(22, 24, 25), and, subsequently, arrest tumor xenograft growth in vivo(24, 25). Recently, a mutant human cyclin G1 (dnG1) was created with a deletion in the cyclin box, a conserved region among cyclins that helps determine cyclin-Cdk association that induces Cdk activation.4Initially shown to exhibit antiproliferative properties in vascular smooth muscle cells, dnG1 may act to inhibit the function of WT cyclin G1 or form inactive complexes with target Cdk molecules. Hence, the performance of a series of cytocidal mutant cyclin G1 constructs were tested in vitro to determine the optimal construct for additional in vivo studies.

A human undifferentiated cancer cell line of pancreatic origin was selected as the prototype of a metastatic gastrointestinal cancer. Retroviral transduction efficiency in these cancer cells was excellent,ranging from 26–85%, depending on the multiplicity of infection(multiplicity of infection = 4 and 250,respectively; Fig. 1,A). For selection of an optimal therapeutic gene, cell proliferation studies were conducted in transduced cells using vectors bearing various cyclin G1 constructs. Fig. 1,B shows the cytocidal/cytostatic effects of mutant and antisense cyclin G1 retroviral vectors in transduced cancer cells. Under standard conditions, the dnG1 vector consistently exhibited the greatest antiproliferative effect, concomitant with the appearance of immunoreactive cyclin G1 at the region of Mr 20,000, representing the dnG1 protein (Fig. 1,C). The expression of the dnG1 protein in cancer cells results in unscheduled cell death; hence, the low-level expression of dnG1 protein in cell lysates from transduced unselected cell cultures. Based on the increased cytocidal activity of the dnG1 vector (Fig. 1 B) in transduced MiaPaca2 cancer cells, the dnG1 vector was used in subsequent in vivoefficacy studies.

Considering the overall mass of the liver, the physiological dilution factors involved, and the practical limitations of vector volume(∼10% of blood volume), the efficiency of matrix-targeted gene delivery by portal vein infusions was remarkably high, ranging from 1–3% (20) to ≥50% in focal areas (Fig. 2). Under these conditions, repeated injections of a high-dose matrix-targeted retroviral vector bearing a β-galactosidase marker gene were required to detect an appreciable gene transfer in the tumor foci. As shown in Fig. 2, enhanced transduction of metastatic tumor cells (≥50%) was observed not only within established tumor nodules as described previously (20) but also at points of tumor exit from the circulation (Fig. 2, A–C). Moreover, significant transduction (≥50%) of tumor-associated stromal cells and endothelial cells was observed in some tumor foci undergoing active angiogenesis (Fig. 2, D–F). No evidence of transduction of neighboring hepatocytes was apparent, indicating the selective retroviral transduction of actively dividing tumor cells.

Histological and immunohistochemical evaluation of metastatic tumor foci from mice treated with either PBS or the low-dose dnG1 vector was performed and evaluated by morphometric analysis with an Optimas imaging system. Histological examination of liver sections from control animals revealed substantial tumor foci with attendant areas of angiogenesis and stroma formation (Fig. 3, A and C); the epithelial components stained positive for cytokeratin, and the associated tumor stromal/endothelial cells stained positive for vimentin and FLK receptor (data not shown). This heterogenous population of tumor cells, stromal cells, and endothelial cells constituted the tumor nodule. In contrast,the mean size of tumor foci in the low-dose dnG1-treated animals was significantly reduced compared with that in PBS controls (Fig. 3, B and D, arrows; P = 0.001), simultaneously revealing a focal increase in the density of apoptotic nuclei (Fig. 3, F and H, arrows) compared with that in the PBS-treated control group (Fig. 3,E). Furthermore, infiltration by PAS+, CD68+, and hemosiderin-laden macrophages (Fig. 3 D, arrow) was observed in the residual tumor foci of dnG1 vector-treated animals, suggesting active clearance of degenerating tumor cells and tumor debris by the hepatic reticuloendothelial system.

Morphometric analysis of tumor foci confirmed that the targeting strategy for therapeutic gene delivery was effective in that portal vein infusions (via indwelling catheter) of high-dose matrix-targeted dnG1 vectors induced dramatic reductions in the sizes of tumor foci when compared with the PBS- and control vector-treated animals based on all response variables (P < 0.0002; Tables 1,2). In pairwise comparisons for the three outcome variables, a dose-dependent tumor response to dnG1 vector treatment was apparent,and additional studies are currently underway to better determine tumor responsiveness to various vector doses in terms of tumor shrinkage versus complete disappearance of the tumor foci and to predict the minimum effective vector dose that could achieve the desired tumor response in Phase I/II gene therapy trials. Importantly,no evidence of hepatocellular damage, necrosis, thrombosis, or cholestasis was detected in tissue sections from dnG1 vector-treated animals, indicating that the matrix-targeted dnG1 vector (cumulative dose,106−109 cfu) may have a wide margin of safety. Whereas our previous studies showed that the level of retroviral transduction of tumor nodules in vivowas only 1–3% (20), the efficiency of gene delivery was dramatically enhanced by increasing the number of vector infusions(Fig. 2). A heterogeneous population of cells within the tumor nodule was transduced, with the greatest number of transduced tumor cells and tumor stromal and endothelial cells at sites of tumor invasion and in areas of active angiogenesis. An appealing concept is that tumor stromal cells and associated endothelial cells were transduced by the dnG1 vector simultaneously with the tumor cells, resulting in disproportionate tumoricidal effects caused by a decrease in vascular supply and/or growth factor stimuli. We have shown previously that blockade of cyclin G1 function using an antisense cyclin G1(22) or a dnG1 vector (data not shown) induces apoptosis in ∼35% of transduced cancer cell cultures. The simultaneous increase in the incidence of apoptotic nuclei in the tumor foci of dnG1 vector-treated mice indicates that dnG1 inhibits tumor growth by evoking apoptotic mechanisms.

In conclusion, we conducted in vivo efficacy and safety studies in a unique nude mouse model of liver metastasis and established the proofs of principle that (a) therapeutic gene delivery can be achieved by repeated portal vein infusions (via an indwelling catheter) of matrix-targeted retroviral vectors bearing a cytocidal mutant cyclin G1 construct, as evidenced by statistically significant reductions in the sizes of tumor foci in dnG1 vector-treated mice compared with those of tumor foci in control animals, and (b) matrix-targeted dnG1 vectors may be systemically administered with a wide margin of safety, as indicated by the absence of associated hepatocyte necrosis, thrombosis, or cholestasis. Taken together, these findings represent a definitive advance toward the development of targeted injectable gene therapy vectors for metastatic cancer.

Fig. 1.

A, transduction efficiency of matrix-targeted retroviral vectors in MiaPaca2 cells.β-galactosidase-expressing cells are shown with blue-stained nuclei. B, cytostatic effects of matrix-targeted retroviral vectors bearing mutant cyclin G1 constructs in MiaPaca cancer cells. The number of cells/well (plotted on the vertical axis)is expressed as a function of time (days after transduction), which is plotted on the horizontal axis. D10, control medium; Null, vector bearing only the neor gene; AS 587 and AS 693, vectors bearing antisense cyclin G1 constructs; dnG1, dnG1 vector bearing a deletion in the NH2 terminus of human cyclin G1. C, Western analysis of human cyclin G1 protein expression in dnG1 vector-transduced versus null vector-transduced cancer cells without G418 selection. Immunoreactive dnG1 (cyclin G1 DN41) is detected as a distinct band in the region of Mr 20,000 (Lane 3), whereas the endogenous cyclin G1 protein is seen as an intensely stained band in the region of Mr ∼30,000 (Lanes 1–4).

Fig. 1.

A, transduction efficiency of matrix-targeted retroviral vectors in MiaPaca2 cells.β-galactosidase-expressing cells are shown with blue-stained nuclei. B, cytostatic effects of matrix-targeted retroviral vectors bearing mutant cyclin G1 constructs in MiaPaca cancer cells. The number of cells/well (plotted on the vertical axis)is expressed as a function of time (days after transduction), which is plotted on the horizontal axis. D10, control medium; Null, vector bearing only the neor gene; AS 587 and AS 693, vectors bearing antisense cyclin G1 constructs; dnG1, dnG1 vector bearing a deletion in the NH2 terminus of human cyclin G1. C, Western analysis of human cyclin G1 protein expression in dnG1 vector-transduced versus null vector-transduced cancer cells without G418 selection. Immunoreactive dnG1 (cyclin G1 DN41) is detected as a distinct band in the region of Mr 20,000 (Lane 3), whereas the endogenous cyclin G1 protein is seen as an intensely stained band in the region of Mr ∼30,000 (Lanes 1–4).

Close modal
Fig. 2.

X-Gal immunostaining for detection of theβ-galactosidase transgene in metastatic tumor nodules. A, H&E- and X-Gal-stained tissue section of liver from aβ-galactosidase vector-treated mouse model of liver metastasis(×40). B, X-Gal-stained tissue section of A counterstained with nuclear fast red stain (×40). C, higher magnification of B (×200).β-Galactosidase-expressing tumor cells (cells with blue-stained nuclei) near a disrupted hepatic venule(v) are indicated by arrows. D, H&E-stained tissue section of liver from aβ-galactosidase vector-treated mouse model of liver metastasis (×40)showing the tumor nodule undergoing active angiogenesis. E, X-Gal-stained liver sections counterstained with nuclear fast red stain (×40). F, higher magnification of E (×200). β-Galactosidase-expressing tumor stromal and endothelial cells are indicated by arrows.

Fig. 2.

X-Gal immunostaining for detection of theβ-galactosidase transgene in metastatic tumor nodules. A, H&E- and X-Gal-stained tissue section of liver from aβ-galactosidase vector-treated mouse model of liver metastasis(×40). B, X-Gal-stained tissue section of A counterstained with nuclear fast red stain (×40). C, higher magnification of B (×200).β-Galactosidase-expressing tumor cells (cells with blue-stained nuclei) near a disrupted hepatic venule(v) are indicated by arrows. D, H&E-stained tissue section of liver from aβ-galactosidase vector-treated mouse model of liver metastasis (×40)showing the tumor nodule undergoing active angiogenesis. E, X-Gal-stained liver sections counterstained with nuclear fast red stain (×40). F, higher magnification of E (×200). β-Galactosidase-expressing tumor stromal and endothelial cells are indicated by arrows.

Close modal
Fig. 3.

H&E-stained tissue sections of liver reveal the tumor foci in the PBS-treated control group (A and C; ×40 and ×100) and the dnG1 vector-treated animals(B and D; ×40 and ×100). Apoptosis in the tumor foci of the PBS-treated control group (E;×100) and the dnG1 vector-treated animals (F and H, ×100 and ×200; arrows) is depicted as reddish-brown immunostaining in an ApopTagPlus peroxidase in situ apoptosis assay. G,negative staining control without the terminal deoxynucleotidyltransferase enzyme; t, tumor foci; H, hepatocytes in liver parenchyma.

Fig. 3.

H&E-stained tissue sections of liver reveal the tumor foci in the PBS-treated control group (A and C; ×40 and ×100) and the dnG1 vector-treated animals(B and D; ×40 and ×100). Apoptosis in the tumor foci of the PBS-treated control group (E;×100) and the dnG1 vector-treated animals (F and H, ×100 and ×200; arrows) is depicted as reddish-brown immunostaining in an ApopTagPlus peroxidase in situ apoptosis assay. G,negative staining control without the terminal deoxynucleotidyltransferase enzyme; t, tumor foci; H, hepatocytes in liver parenchyma.

Close modal

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1

Supported in part by a grant from the Whittier Family Foundation (Pasadena, CA) and by National Cancer Institute/NIH Grant P30CA14089.

3

The abbreviations used are: ECM, extracellular matrix; VSVG, vesicular stomatitis virus G; WT, wild-type; vWF, von Willebrand factor; cfu, colony-forming unit(s); SA, surface area; Cdk,cyclin-dependent kinase.

4

Unpublished data.

Table 1

Inhibition of liver metastatic tumors after portal vein infusions of matrix-targeted retroviral vectors bearing a mutant cyclin G1 construct

VariableaTreatment groupnMean95% CIbP
Liver SA PBS 87.72 (56.915–135.204) 0.95 
 High-dose nBg 89.85 (58.299–138.491)  
 Low-dose dnG1 101.41 (65.795–156.300)  
 High-dose dnG1 97.95 (63.550–150.967)  
 High-dose nBg:PBS ratio  1.02 (0.556–1.889)  
 Low-dose dnG1:PBS ratio  1.16 (0.627–2.131)  
 High-dose dnG1:PBS ratio  1.12 (0.606–2.059)  
Tumor SA PBS 0.847 (0.210–3.418) 0.006 
 High-dose nBg 0.188 (0.047–0.760)  
 Low-dose dnG1 0.268 (0.066–1.083)  
 High-dose dnG1 0.016 (0.004–0.063)  
 High-dose nBg:PBS ratio  0.222 (0.031–1.600)  
 Low-dose dnG1:PBS ratio  0.317 (0.044–2.280)  
 High-dose dnG1:PBS ratio  0.018 (0.003–0.133)  
Tumor SA: liver SA× 100 PBS 0.965 (0.224–4.155) 0.007 
 High-dose nBg 0.210 (0.049–0.902)  
 Low-dose dnG1 0.265 (0.061–1.139)  
 High-dose dnG1 0.016 (0.004–0.069)  
 High-dose nBg:PBS ratio  0.217 (0.028–1.711)  
 Low-dose dnG1:PBS ratio  0.274 (0.035–2.160)  
 High-dose dnG1:PBS ratio  0.017 (0.002–0.130)  
Mean tumor SA PBS 0.109 (0.054–0.218) 0.0002 
 High-dose nBg 0.042 (0.021–0.085)  
 Low-dose dnG1 0.012 (0.006–0.023)  
 High-dose dnG1 0.006 (0.003–0.013)  
 High-dose nBg:PBS ratio  0.389 (0.145–1.039)  
 Low-dose dnG1:PBS ratio  0.106 (0.040–0.284)  
 High-dose dnG1:PBS ratio  0.059 (0.022–0.158)  
VariableaTreatment groupnMean95% CIbP
Liver SA PBS 87.72 (56.915–135.204) 0.95 
 High-dose nBg 89.85 (58.299–138.491)  
 Low-dose dnG1 101.41 (65.795–156.300)  
 High-dose dnG1 97.95 (63.550–150.967)  
 High-dose nBg:PBS ratio  1.02 (0.556–1.889)  
 Low-dose dnG1:PBS ratio  1.16 (0.627–2.131)  
 High-dose dnG1:PBS ratio  1.12 (0.606–2.059)  
Tumor SA PBS 0.847 (0.210–3.418) 0.006 
 High-dose nBg 0.188 (0.047–0.760)  
 Low-dose dnG1 0.268 (0.066–1.083)  
 High-dose dnG1 0.016 (0.004–0.063)  
 High-dose nBg:PBS ratio  0.222 (0.031–1.600)  
 Low-dose dnG1:PBS ratio  0.317 (0.044–2.280)  
 High-dose dnG1:PBS ratio  0.018 (0.003–0.133)  
Tumor SA: liver SA× 100 PBS 0.965 (0.224–4.155) 0.007 
 High-dose nBg 0.210 (0.049–0.902)  
 Low-dose dnG1 0.265 (0.061–1.139)  
 High-dose dnG1 0.016 (0.004–0.069)  
 High-dose nBg:PBS ratio  0.217 (0.028–1.711)  
 Low-dose dnG1:PBS ratio  0.274 (0.035–2.160)  
 High-dose dnG1:PBS ratio  0.017 (0.002–0.130)  
Mean tumor SA PBS 0.109 (0.054–0.218) 0.0002 
 High-dose nBg 0.042 (0.021–0.085)  
 Low-dose dnG1 0.012 (0.006–0.023)  
 High-dose dnG1 0.006 (0.003–0.013)  
 High-dose nBg:PBS ratio  0.389 (0.145–1.039)  
 Low-dose dnG1:PBS ratio  0.106 (0.040–0.284)  
 High-dose dnG1:PBS ratio  0.059 (0.022–0.158)  
a

Liver SA, total surface area of liver section, mm2 (mean of lobes A, B, and C); Tumor SA,total surface area of tumor foci per lobe, mm2 (mean of lobes A, B, and C); Tumor SA: liver SA × 100, percentage of liver lobe infiltrated by tumor; Mean Tumor SA, mean surface area of each tumor foci in lobes A, B, and C, mm2.

b

CI, confidence interval.

Table 2

Pairwise comparisons of the outcome variables in Table 1 a

Tumor sizeTumor:liver size ratioMean tumor size
High-dose nBg vs. PBS 0.123 0.133 0.058 
Low-dose dnG1 vs. PBS 0.229 0.197 <0.001 
High-dose dnG1 vs. PBS 0.001 0.001 <0.001 
High-dose nBg vs. low-dose dnG1 0.703 0.810 0.014 
High-dose nBg vs. high-dose dnG1 0.018 0.019 0.001 
Low-dose dnG1 vs. high-dose dnG1 0.009 0.012 0.220 
Tumor sizeTumor:liver size ratioMean tumor size
High-dose nBg vs. PBS 0.123 0.133 0.058 
Low-dose dnG1 vs. PBS 0.229 0.197 <0.001 
High-dose dnG1 vs. PBS 0.001 0.001 <0.001 
High-dose nBg vs. low-dose dnG1 0.703 0.810 0.014 
High-dose nBg vs. high-dose dnG1 0.018 0.019 0.001 
Low-dose dnG1 vs. high-dose dnG1 0.009 0.012 0.220 
a

Pairwise comparisons were performed for the outcome variables with overall F test; Ps were <0.05 between groups.

We thank Drs. Fan Xu and Liqiong Liu for technical assistance in cloning of the dnG1 vector and the retroviral vector targeting constructs.

1
Anderson W. F. Human gene therapy.
Nature (Lond.)
,
392(Suppl.)
:
25
-30,  
1998
.
2
Gordon E. M., Anderson W. F. Gene therapy using retroviral vectors.
Curr. Opin. Biotechnol.
,
5
:
611
-616,  
1994
.
3
Miller D. G., Adam M. A., Miller A. D. Gene transfer by retrovirus vectors occurs only in cells that are actively replicating at the time of infection.
Mol. Cell. Biol.
,
10
:
4239
-4242,  
1990
.
4
Hunter E., Swanstrom R. Retrovirus envelope glycoproteins.
Curr. Top. Microbiol. Immunol.
,
157
:
187
-253,  
1992
.
5
Weiss R. A. Cellular receptors and viral glycoproteins involved in retrovirus entry Levy J. A. eds. .
The Retroviridae
,
:
1
-108, Plenum Press New York  
1993
.
6
White J. M. Membrane fusion.
Science (Washington DC)
,
258
:
917
-924,  
1992
.
7
Soneoka Y., Cannon P. M., Ramsdale E. E., Griffiths J. C., Gaetano R., Kingsman S. M., Kingsman A. J. A transient three-plasmid expression system for the production of high titer retroviral vectors.
Nucleic Acids Res.
,
23
:
628
-633,  
1995
.
8
Yang S., Delgado R., King S. R., Woffendin C., Barker C. S., Yang Z. Y., Xu L., Nolan G. P., Nabel G. J. Generation of retroviral vector for clinical studies using transient transfection.
Hum. Gene Ther.
,
10
:
123
-132,  
1999
.
9
Pensiero M. N., Wysocki C. A., Nader K., Kikuchi G. E. Development of amphotropic retrovirus vectors resistant to inactivation by human serum.
Hum. Gene Ther.
,
8
:
1095
-1101,  
1996
.
10
Verma I., Somia N. Gene therapy: promises, problems and prospects.
Nature (Lond.)
,
389
:
239
-242,  
1997
.
11
Kasahara N., Dozy A. M., Kan Y. W. Tissue-specific targeting of retroviral vectors through ligand-receptor interactions.
Science (Washington DC)
,
266
:
1373
-1376,  
1994
.
12
Cosset F. L., Russell S. J. Targeting retrovirus entry.
Gene Ther.
,
3
:
946
-956,  
1996
.
13
Peng K. W., Morling F. J., Cosset F. L., Murphy G., Russell S. J. A gene delivery system activatable by disease-associated matrix metalloproteinases.
Hum. Gene Ther.
,
8
:
729
-738,  
1997
.
14
Valsesia-Wittman S., Drynda A., Deleage G., Aumailley M., Heard J. M., Danos O., Verdier G., Cosset F-L. Modification in the binding domain of avian retrovirus envelope protein to redirect the host range of retroviral vectors.
J. Virol.
,
68
:
4609
-4619,  
1994
.
15
Russell S. J. Retroviral vectors displaying functional antibody fragments.
Nucleic Acids Res.
,
21
:
1081
-1085,  
1993
.
16
Cosset F. L., Morling F. J., Takeuchi U., Weiss R. A., Collins M. K., Russell S. J. Retroviral targeting by envelopes expressing an N-terminal binding domain.
J. Virol.
,
69
:
6314
-6322,  
1995
.
17
Somia N. V., Zoppe M., Verma I. M. Generation of targeted retroviral vectors by using single-chain variable fragment: an approach to in vivo gene delivery.
Proc. Natl. Acad. Sci. USA
,
92
:
7570
-7574,  
1995
.
18
Hall F. L., Gordon E. M., Wu L., Zhu N. L., Skotzko M. J., Starnes V. A., Anderson W. F. Targeting retroviral vectors to vascular lesions by genetic engineering of the MoMuLV gp70 envelope protein.
Hum. Gene Ther.
,
8
:
2183
-2192,  
1997
.
19
Wu B. W., Cannon P. M., Gordon E. M., Hall F. L., Anderson W. F. Characterization of the proline-rich region of murine leukemia virus envelope protein.
J. Virol.
,
72
:
5383
-5391,  
1998
.
20
Hall F. L., Liu L., Zhu N. L., Stapfer M., Anderson W. F., Beart R. W., Gordon E. M. Molecular engineering of matrix-targeted retroviral vectors incorporating a surveillance function inherent in von Willebrand factor.
Hum. Gene Ther.
,
11
:
983
-993,  
2000
.
21
Yee T. K., Friedmann T., Burns J. C. Generation of high-titer pseudotyped retroviral vectors with very broad host range.
Methods Cell Biol.
,
43
:
99
-112,  
1994
.
22
Skotzko M. J., Wu L. T., Anderson W. F., Gordon E. M., Hall F. L. Retroviral vector-mediated gene transfer of antisense cyclin G1 (CYCG1) inhibits proliferation of human osteogenic sarcoma cells.
Cancer Res.
,
55
:
5493
-5498,  
1995
.
23
Wu L., Liu L., Yee A., Carbonaro-Hall D., Tolo V. T., Hall F. L. Molecular cloning of the human CYCG1 gene encoding a G-type cyclin: overexpression in osteosarcoma cells.
Oncol. Rep.
,
1
:
705
-711,  
1994
.
24
Chen D. S., Zhu N. L., Hung G., Skotzko M. J., Hinton D., Tolo V., Hall F. L., Anderson W. F., Gordon E. M. Retroviral vector-mediated transfer of an antisense cyclin G1 construct inhibits osteosarcoma tumor growth in nude mice.
Hum. Gene Ther.
,
8
:
1679
-1686,  
1997
.
25
Hung G., Skotzko M. J., Chang M., Zhu N. L., Parekh D., Hall F. L., Gordon E. M., Anderson W. F. Intratumoral injection of an antisense cyclin G1 retroviral vector inhibits growth of undifferentiated carcinoma xenografts in nude mice.
Int. J. Pediatr. Hematol. Oncol.
,
4
:
317
-325,  
1997
.